Novel Technology May Detect Early Warning Signs of COVID-19 Exacerbation

Originally published in Chicago Medicine Magazine

Continuous RPM + AI personalized analytics show promise helping providers manage patient care

By: Gary Manning, Senior Vice President and General Manager, Healthcare, physIQ, and

Karen Larimer, PhD, ACNP-BC, Director, Clinical Development, physIQ

During the past year, we learned a great deal about how SARS-CoV-2, the virus that causes COVID-19, is transmitted. We also learned that about one in five people who contract COVID-19 are at risk for severe decompensation that results in hospitalization.

Yet we still know little about the factors that predict a symptom-free or mild disease course versus a severe one, making it challenging for healthcare providers to determine which patients will decompensate (or worsen) and require intervention. This information is especially crucial given that the earlier in the disease progression COVID-19 patients are treated, the more likely they are to have better outcomes.

Preliminary research indicates that an innovative digital health technology approach has the potential to deliver valuable insights about patient decompensation risk. A peer-reviewed article recently published in npj Digital Medicine documented the development and validation of a first-of-its-kind COVID-19 Decompensation Index (CDI), based on continuous remote monitoring of 400 adults in Phase I of a two-phase clinical study. The study included patients who tested positive for SARS-CoV-2 in an outpatient setting and those who had been discharged following hospitalization with the disease. They were followed remotely to assess when and if they developed COVID-19.

The NIH-funded DeCODe study sought to develop and validate a CDI using data derived from wearable biosensors. An FDA-cleared artificial intelligence (AI) analytics platform (pinpointIQ™ from physIQ) continuously monitored patient vital signs through a chest-worn biosensor. AI-driven personalized analytics were applied to the sensor-derived multi-parameter data and alerted clinicians of any notable physiologic changes. Based on this information and subsequent telephone conversation with the patient, clinicians determined if intervention was required.

The machine learning CDI model outperformed standard of care (SOC) modalities, identifying early warning signs of potentially worsening health more frequently.

The study and these initial findings are promising for the following reasons:

1. Fully remote clinical study deployment works – and delivers multiple benefits.

The DeCODe study demonstrated that it was possible to quickly and effectively deploy a remote monitoring system with no in-person engagement. This was critical for the comfort of the patient and safety of the clinical team. The entire enrollment and monitoring process was handled remotely, including requesting and receiving patient consent, sending the sensor kit, and educating patients about how to properly wear the biosensor and monitor their vital signs.

Patient adherence to clinical study protocols always presents challenges, even when patients receive face-to-face instruction, let alone in a fully remote deployment. Would patients wear a biosensor, properly and for the required amount of time? Would they provide information and complete surveys as requested? The answers turned out to be a resounding “yes.” Patients wore the biosensor more than 90 percent of the time, and more than 85 percent of patients answered the survey questions as scheduled.

The remote approach and the simplicity of the system also delivered vital benefits. First and foremost, it supported safety protocols during the height of the pandemic, protecting patients and providers and enabling the study to proceed even when most others were suspended.

It also opened the door to recruiting a wider pool of patients, geographically, economically and demographically. Eliminating the need for patients to come to a particular location in person means they can live anywhere and do not need to arrange or pay for transportation. In addition, using a cellular rather than WiFi based system makes it possible for patients in rural or underserved metro areas to participate.

2. Accelerated clinical study execution supports innovation.

Funded as part of Operation Warp Speed, the DeCODe study was executed at an unprecedented pace. Within two-and-a-half months, the contract was funded and 400 patients were enrolled, a process that often takes years. The ability to ramp up quickly during an especially challenging time and successfully demonstrate the feasibility – and viable scalability – of the continuous RPM system for detecting potential exacerbation of COVID-19 delivered real value. This expedited approach of “succeed fast or fail fast” is key to spurring innovation.

3. The CDI is being built on data from a diverse population.

The Phase I study enrolled 400 adults diagnosed with COVID-19 from the University of Illinois Health System (UI Health). Through its federally qualified health centers as well as its in-patient hospitals, UI Health was able to recruit patients from underserved areas hardest hit by COVID-19, aligning the study with the system’s mission to engage diverse communities and advance health equity.

As a result, the CDI was built on data from a population that was 46 percent Hispanic and 37 percent non-Hispanic Black. Diverse representation is important to demonstrating the CDI’s applicability for monitoring, treating and potentially improving health outcomes for all patients. These data also played a valuable role in meeting an additional objective of the study – contributing to a national digital data hub to support ongoing COVID-19 research.

4. Near real-time data leads to earlier clinical intervention.

Relying on traditional spot-check physiological measurements, which are often lagging indicators or inconsistent, can limit providers’ ability to determine appropriate treatment. This is especially true in the case of COVID-19, which manifests with a variety of symptoms. In contrast, combining wearable biosensors with machine learning-based analytics generates a continuous stream of clean, high-quality data that essentially functions as an early warning system.

Clinicians are alerted to any notable physiologic changes so they can follow-up right away with a call to the patient. While engaging with the patient remains crucial, having the respiratory rate and heart rate data supplements patient self-assessment and provides physicians and nurses with a more robust picture of exactly what’s happening. In cases where patients appear to be at risk of deterioration, clinicians can intervene early and promptly to reduce the risk of complications.

5. Continuous RPM reassures patients and minimizes fear.

When the study began, best practices for COVID-19 management were still being refined, and patients in outpatient settings expressed significant fear about their health. Many had heard stories about patients who remained at home and believed they were getting better but instead became gravely ill. For COVID-19 positive patients at home, knowing that a clinician routinely viewed their vital signs and could take action as needed, such as sending them to urgent care or the hospital, comforted and helped them feel safe. In fact, many expressed reluctance about giving up the monitoring after the prescribed 28 days.

6. Clinicians can readily incorporate digital technology into their workflow.

The team of nurse practitioners involved in the study initially checked the clinical portal three times during the day. Quickly, however, they discovered they could trust the data and the analytics and began checking only once daily. They also were able to incorporate the system smoothly into their regular patient management workflow and begin to envision how it could be used to guide care decision-making for other diseases as well as COVID-19.

7. Identifying patients most at risk helps allocate limited healthcare resources.

With the ongoing shortage of physicians, nurses and other clinicians, the ability to determine which patients need higher levels of care – and when and where – enables healthcare systems to allocate staff, equipment and other resources most effectively. Technology like this system makes it possible for clinicians to confidently monitor scores of patients at one time, recognizing they may only need to actively manage a much smaller number based on the alerts they receive.

What’s next?

Phase II of the DeCODe study enrolled an additional 600 COVID-positive patients through UI Health, NorthShore University HealthSystem, Rush University Medical Center, Intermountain Healthcare and the University of Texas Houston Medical Center. Initial results have been reported to the National Institutes of Health.

In addition, physIQ is continuing to partner with UI Health to use pinpointIQ to monitor 40 patients with post-acute SARS-CoV-2 syndrome, the long-haulers who remain symptomatic for extended periods even after they have technically recovered from the disease.

Most importantly, as providers and patients alike demonstrate an eagerness to embrace virtual care models, the DeCODe study has demonstrated the feasibility of using continuous RPM and AI analytics to detect exacerbation of COVID-19 in a diverse patient population. The challenge now is to continue to explore how novel digital health technologies can help deepen our understanding of disease, enhance care delivery and support better health outcomes for all patients.